Classification of CMT Additive Manufacturing Defects based on Statistical Features

Abstract

Aiming at the problem of the CMT(cold metal transfer) additive manufacturing process defect monitoring method being susceptible to environmental influence, this paper proposes a classification method for additive manufacturing process defects with current as the research object. By setting up three sets of experiments, experimenting and collecting the current parameters of each group, using the statistical feature method to pretreat the current, and then classifying the pretreated data into random forests, to highlight the superiority of statistical characteristics plus random forests, two sets of control groups are also set up for comparison. It was concluded that the principal component analysis method could not classify the non-defective group and the non-protective gas group, while the use of random forest could achieve two by two classifications of the three working conditions; After the statistical feature method treatment, the random forest classification is compared with the original current direct random forest classification, and the accuracy of the training set, verification set, and test set has been improved.